Identification of diseases by their scent has the potential to revolutionize diagnostics. We believe that many diseases have their unique scent, which can be identified by non-invasive, cheap methods. Each scent is determined by unique patterns of specific volatile organic compounds (VOC). We have developed an electronic nose (or e-nose) that can identify and discriminate different VOC pattens with high accuracy. In this proof-of-concept study to be implemented in Botswana, we will focus on the identification of M. tuberculosis (MTB) through the recognition of is VOC patterns under laboratory and field conditions. First, we will determine the sensitivity and specificity of the e-nose on cultures and sputum samples with known concentrations of MTB, as well as on clinical sputum samples from patients with confirmed pulmonary tuberculosis (PTB) under laboratory, controlled conditions. Then, we will test the performance of the e-nose on the identification of MTB on the breath of patients with active PTB. These data will be used to identify the VOC patterns that best screen and diagnose MTB under laboratory and field conditions, aimed at assisting the non-specialist health care provider in clinical decision making. Since the only requirement for an e-nose i the partial selectivity of the sensors, it is possible to assemble electronic noses withany available sensor technology. This study will be the first one to comprehensively determine the relative performance of a wide variety of sensors and of e-noses platforms build with different number of such sensors under different environmental conditions.